Azure Bing Grounding

Web search grounding via Azure Foundry and Bing Grounding Search tool. Use when the user needs up-to-date information searched from the web via Azure AI Agen...

Security Scan

The skill's code and instructions appear to implement Azure/Bing grounding as described, but the registry metadata omits required environment variables and there are a few operational details you should verify before installing and running (notably Azure credentials and package installation).

Purpose & Capability

The SKILL.md, README, and included Python script all implement an Azure AI Agents + Bing Grounding workflow that matches the skill name/description. However the registry metadata claims no required env vars or primary credential while the SKILL.md and script clearly require FOUNDRY_PROJECT_ENDPOINT and BING_PROJECT_CONNECTION_ID (and optionally AZURE_* credentials). This metadata omission is an inconsistency.

Instruction Scope

Runtime instructions are narrowly scoped to invoking the included script, installing the Azure SDKs, and providing Azure/Azure Foundry/Bing Grounding configuration. The script reads ~/.openclaw/.env and environment variables to obtain configuration/credentials, which is expected for this integration.

Install Mechanism

There is no formal install spec in the registry (instruction-only style), but the SKILL.md requires pip installing azure-identity and azure-ai-agents. These are public Python packages (no direct downloads from arbitrary URLs), so installation risk is moderate and traceable — you should verify package names/versions from PyPI if concerned.

Credentials

The script legitimately needs Azure/Founry/Bing connection information and possibly service-principal secrets to authenticate. That said, the registry metadata does not declare these required env vars (FOUNDRY_PROJECT_ENDPOINT, BING_PROJECT_CONNECTION_ID, AZURE_TENANT_ID, AZURE_CLIENT_ID, AZURE_CLIENT_SECRET), causing a mismatch. Providing Azure credentials grants this skill access to your Azure resources — ensure least-privilege credentials are used.

Persistence & Privilege

The skill is not always-enabled and does not request any special platform-level persistence. It creates and deletes an agent in your Azure project for each run (the script attempts cleanup), which is consistent with its purpose.

What to consider before installing

This skill appears to implement exactly what it claims (Azure AI Agents + Bing Grounding), but the registry metadata failed to declare required environment variables. Before installing or running: 1) Confirm you are comfortable providing FOUNDRY_PROJECT_ENDPOINT and BING_PROJECT_CONNECTION_ID and, if needed, a service principal (AZURE_TENANT_ID, AZURE_CLIENT_ID, AZURE_CLIENT_SECRET). Use a least-privilege service principal. 2) Verify the pip packages (azure-identity, azure-ai-agents) from PyPI and consider pinning versions. 3) Be aware the script reads ~/.openclaw/.env and environment variables — keep only necessary secrets there. 4) Review Azure permissions and billing implications: the script creates agents/operations in your project. 5) If you cannot trust the source, do not provide long‑lived credentials; consider creating a scoped, revocable service principal for testing. If you want me to, I can (a) list the exact env vars the script reads, (b) produce a minimal example service principal permission set, or (c help verify the PyPI packages and current package maintainers.

Use the bundled Python script to perform grounded searches using Azure Foundry's Agent Service and the Bing Grounding Search Tool.

Requirements

1. Required Python packages:

   pip install azure-identity azure-ai-agents
   

2. Authentication:

   • Ensure Azure CLI is logged in (az login), OR
   • Set Azure Service Principal / Managed Identity credentials compatible with DefaultAzureCredential or ClientSecretCredential.

3. Environment Variables: Add the following to your ~/.openclaw/.env file or export them in your shell:

   # Your Azure AI Foundry project endpoint
   FOUNDRY_PROJECT_ENDPOINT="https://<your-resource>.ai.azure.com/api/projects/<your-project>"
   
   # The ID of the Bing Grounding connection in your Azure AI Foundry Project
   BING_PROJECT_CONNECTION_ID="<your-connection-id>"
   
   # Default model deployment name (optional, defaults to gpt-4o)
   FOUNDRY_MODEL_DEPLOYMENT_NAME="gpt-4o"
   
   # (Optional) Service Principal Credentials if not using DefaultAzureCredential
   AZURE_TENANT_ID="<tenant-id>"
   AZURE_CLIENT_ID="<client-id>"
   AZURE_CLIENT_SECRET="<client-secret>"
   

Commands

Run from the OpenClaw workspace:

# Raw JSON output (default)
python3 {baseDir}/scripts/bing_grounding.py --query "What is the latest AI news today?"

# Markdown human-readable output
python3 {baseDir}/scripts/bing_grounding.py --query "What is the latest AI news today?" --format md

# Use a specific model deployment (default is gpt-4o)
python3 {baseDir}/scripts/bing_grounding.py --query "Weather in Seattle?" --model "gpt-4o-mini"

Output

Returns a generated response synthesized by the Azure AI Agent based on Bing Search results, along with the source URL citations.